A salt-tolerant variety identification method based on the combination of multiple indicators and multispectral texture

Abstract

The invention discloses a salt-tolerant variety identification method based on the combination of multiple indicators and multispectral texture, and specifically relates to the field of variety identification. The invention uses a multifunctional plant photosynthetic phenotype measurement system PlantExplorer Pro+ to perform comprehensive measurement of fluorescence quenching kinetic curves of plants to be tested; uses the Pearson correlation coefficient to analyze the correlation between indicators; uses a principal component analysis method to reduce the dimension of data, combines a PCA result with an indicator weight to calculate a comprehensive salt tolerance coefficient through a membership function, uses VideometerLab 4 to collect multispectral images, uses MATLAB to perform threshold segmentation on the multispectral images, extracts multispectral images of regions of interest, uses an OTSU algorithm to realize image binarization, obtains a binarized image, and calculates the spectral average reflectance of a region in the binarized image where the pixel value is not 0; finally, 1D-CNN and 2D-CNN based on multispectral data of an AlexNet framework are constructed, the two are combined, and a dual-input multimodal convolutional neural network model is proposed.

Description

Technical Field

[0001] The present invention relates to the technical field of variety identification, and more specifically, to a method for identifying salt-tolerant varieties based on the combination of multiple indicators and multi-spectral textures. Background Art

[0002] Salt is one of the main factors limiting plant growth and development. Its excessive presence in the soil will have a large number of negative effects on plant growth and development, or even directly cause death. Therefore, screening salt-tolerant varieties is very important for increasing crop production.

[0003] The current mainstream method for detecting salt stress is to quantify salt stress by quantifying physiological indicators, such as detecting a series of physiological and biochemical indicators, detecting changes in ion content, and recording plant appearance data.

[0004] However, there are still some shortcomings in actual use. For example, these detection methods are time-consuming, costly, require large amounts of human resources, and are inefficient. Secondly, in the analysis stage after obtaining the data, there are also shortcomings such as fixed processes and backward technology. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method for identifying salt-tolerant varieties based on a combination of multiple indicators and multi-spectral textures to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] Step A1: Use the multifunctional plant photosynthetic phenotype measurement system PAM Time protocol measurement program to measure the complete fluorescence quenching kinetic curve, and use the Pearson correlation coefficient to analyze the correlation between fluorescence data;

[0008] Step A2: Use principal component analysis to reduce the dimension of the fluorescence data and transform the fluorescence index into a set of linearly unrelated comprehensive indexes;

[0009] Step A3: Establishing a membership function of the comprehensive salt tolerance coefficient by combining the extracted principal components and index weights to obtain the comprehensive salt tolerance coefficient;

[0010] Step A4: Use MATLAB 2023b software and 4 devices to interface and use Library functions to obtain multispectral image data;

[0011] Step A5: Select the near-infrared channel image and perform grayscale processing, and further apply the threshold segmentation method to extract the region of interest; use the OTSU algorithm to convert the extracted ROI region into a binary image;

[0012] Step A6: For the same ROI area, multiply the binarized image obtained in the previous step with the original multispectral image of each channel to generate a new multispectral image of the area;

[0013] Step A7: Calculate the average spectral reflectance of the non-zero area of ​​the pixel points in the multispectral image and the corresponding texture characteristics; use the gray level co-occurrence matrix method to analyze the gray level connection characteristics between the pixels on the leaf surface;

[0014] Step A8: using the gray-level co-occurrence matrix and the local binary method to extract the near-infrared channel grayscale image with obvious characteristics to calculate the texture characteristics of each strain, using the D value as the dependent variable, establishing a multivariate stepwise regression model with the multispectral band parameter as the independent variable, and constructing the optimal regression equation; wherein the multispectral band parameter is the reflectance data of the wavelength of light;

[0015] Step A9: Use the multispectral data collected by VideometerLab4 to construct a Hybrid network model of the framework.

[0016] Preferably, in step A1, the step of obtaining a complete fluorescence quenching kinetic curve is specifically as follows:

[0017] Step A11: preparing a sample of Chinese cabbage containing a fluorescent marker to be tested;

[0018] Step A12: measuring the fluorescence index of the sample using the multifunctional photosynthetic phenotyping system;

[0019] Step A13: Import the raw data measured by the instrument into the Data Analysis software to extract data and images;

[0020] Preferably, in step A2, the maximum and minimum values ​​in the biological replicates of the control group are removed and the average value is calculated, and the data of the experimental group is compared with the data of the control group to eliminate the differences between varieties. The calculation method of the principal component analysis is specifically as follows:

[0021] There are m principal component parameters obtained by the principal component, and the degree of explanation of the nth principal component for all indicators is k.

[0022] The calculation method for the ratio of the explanation degree of each principal component to the explanation degree of all principal components for all indicators is:

[0023] .

[0024] Preferably, , where S represents the membership function, It is expressed as the score of the cth principal component of the dth accession, is expressed as the minimum value of the cth component, It is represented as the maximum value of the cth component, and n is represented as the nth principal component;

[0025] The calculation method of comprehensive salt tolerance coefficient is:

[0026] (n) = μ(n)×S(n) (value of the dth accession after standardization)

[0027] , where D represents the comprehensive salt tolerance coefficient and n represents the nth principal component.

[0028] Preferably, in step A4, the step of acquiring multispectral image data is specifically as follows:

[0029] Step A41: 4. The equipment is installed and set up correctly, and MATLAB2023b has been installed on the computer; 4 API or SDK is compatible with MATLAB and tools are available;

[0030] Step A42: Install MATLAB's Image Processing Toolbox;

[0031] Step A43: Use 4. Built-in msi interface; Place all files in a directory called "Multispectral Imaging Toolbox"; Add this directory to the Matlab search path by selecting "Set Path..." in the Matlab menu; Type help msi in the MATLAB command box to get an overview of the functions in the toolbox;

[0032] Step A44: Use the msi function to initialize and 4; set the exposure time and spectral range for multispectral image acquisition; call the msi function to obtain multispectral image data;

[0033] Step A45: Pass The library's functions process, analyze and visualize 4. Acquired data.

[0034] Step A46: Use MATLAB's image processing toolbox to perform image denoising and correction; Library for image analysis,

[0035] Step A47: Visualize the results through the plotting function of MATLAB to display the processed multispectral image; export the processed data as a file.

[0036] Preferably, the method for extracting the region of interest is specifically:

[0037] Step A51: Select the grayscale image of the near infrared channel 850nm as the target grayscale image;

[0038] Step A52: selecting a threshold to distinguish the target and background in the image;

[0039] The thresholds selected include manually selected fixed values ​​and values ​​automatically determined based on the histogram or other features of the image;

[0040] Common selection methods include Methods,Histogram-based methods;

[0041] Step A53: applying the selected threshold to the grayscale image, dividing the image into two parts: pixels greater than the threshold are targets, and pixels less than the threshold are backgrounds, thereby obtaining a binary image in which the target pixel value is 1 and the background pixel value is 0, and marking the target image as an image of interest;

[0042] Step A54: Perform some post-processing operations on the binary image, use morphological operations to remove noise or connect separated targets; display the processed binary image to check and analyze the segmentation results.

[0043] Preferably, in step A6, the calculation method for generating a new multispectral image is specifically:

[0044] , where Z represents the multispectral image value of ROI, Represents the reflectivity of each pixel of the original image. represents the value of each pixel of the binary image, N(k) represents the spectral image of the kth band, and the 19 images are superimposed to obtain a 19-channel ROI multispectral image.

[0045] Preferably, in step A7, the method for calculating the average spectral reflectance is specifically as follows:

[0046] ,in, Expressed as the average spectral reflectance, It is represented as the reflectance of the pixel in the i-th row and j-th column in the n-th band, q is represented as the number of non-zero pixels inside the selected ROI, and m and n represent a total of m rows and n columns respectively.

[0047] Preferably, in step A8, the calculation method of multi-spectrum combined with texture in the regression equation is specifically:

[0048] ;

[0049] The calculation method of multispectral data in the regression equation is as follows:

[0050] ;

[0051] in, It is related to the chlorophyll content of cabbage. and Related to chlorophyll content, leaf structure and leaf thickness, Related to the water content of leaves, It is related to the pigment content in plant leaves. LBP indicates the concavity and convexity of the cabbage leaf surface and the roughness of the texture. CONTRAST indicates the contrast of the image.

[0052] Preferably, in step A9, constructing The hybrid network model of the framework is specifically: including a 1D convolutional neural network and a 2D convolutional neural network, and fusing the features of the two together through a splicing layer; specifically, the 1D-CNN architecture is designed to include a convolutional layer, a normalization layer, and a The activation function layer, a downsampling layer, and an output layer; the 2D convolutional neural network structure includes two convolutional layers, two normalization layers, and a downsampling layer, and also cooperates with activation function, and also has an output layer; then, three fully connected layers are added on top of the fused features, and an activation function is set after them as layer and an output layer for classification tasks to distinguish the target categories; finally, in order to evaluate the performance of the 2D convolutional neural network and splicing network model built based on multispectral images, a five-fold cross-validation strategy was adopted. The accuracy rate and key evaluation indicators F1 value, Precision and Recall obtained by five cross-validations were averaged to obtain the overall performance of the model as the final result measurement standard.

[0053] Technical effects and advantages of the present invention:

[0054] 1. The present invention combines the photosynthetic capacity of plants and effectively improves the accuracy and reliability of the identified salt-tolerant varieties through scientific measurement and analysis methods, thereby increasing the credibility of the identification results; by adopting efficient measurement technology and data analysis methods, the present invention significantly reduces the time required in the data collection stage, thereby accelerating the overall process of salt-tolerant variety identification;

[0055] 2. The convolutional layers, normalization layers, and activation functions in 1D-CNN and 2D-CNN work together to reduce the risk of overfitting and enhance the generalization and robustness of the model. In the 1D-CNN and 2D-CNN models used in this patent, the combination of carefully designed convolutional layers, normalization layers, and activation functions reduces the possibility of model overfitting, which helps improve the generalization ability of the model on new samples and also enhances the robustness of the model to potential noise interference in the data;

[0056] 3. The 1D-CNN and 2D-CNN structures are relatively independent and suitable for parallel processing. With the corresponding hardware support, these two parts of the calculation can be performed in parallel, which greatly improves the data processing speed. In addition, the 1D-CNN and 2D-CNN structures adopt a modular design principle, which improves the convenience of engineering implementation and facilitates flexible adjustment and optimization and upgrading according to needs in the later stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a schematic diagram of the process of the present invention.

[0058] Figure 2 The figure is a schematic diagram comparing the spectral reflectance of the control group and the test group of the sensitive Chinese cabbage of the present invention on different days.

[0059] Figure 3 The figure is a schematic diagram comparing the spectral reflectance of the control group and the test group of the salt-tolerant Chinese cabbage of the present invention on different days.

[0060] Figure 4 This is a schematic diagram comparing the accuracy of the MMCNN of the present invention and the ordinary 2D-CNN.

[0061] Figure 5 It is a visual schematic diagram of the salt tolerance grade screening of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] See also Figure 1 As shown, the present invention provides Figure 1 As shown, the present invention provides a salt-tolerant variety identification method based on a combination of multiple indicators and multi-spectral texture, comprising the following steps

[0064] Step A1: Use the multifunctional plant photosynthetic phenotype measurement system PAM Time protocol measurement program to measure the complete fluorescence quenching kinetic curve, and use the Pearson correlation coefficient (PCC) to analyze the correlation between fluorescence data;

[0065] In step A1, the steps of obtaining a complete fluorescence quenching kinetic curve are specifically as follows:

[0066] Step A12: measuring the fluorescence index of the sample using the multifunctional photosynthetic phenotyping system;

[0067] Step A13: Import the raw data measured by the instrument into the Data Analysis software to extract data and images;

[0068] The calculation method of Pearson correlation coefficient is:

[0069] , where P represents the Pearson correlation coefficient, x and y represent different fluorescence characteristics, Expressed as the standard deviation of the index x, Expressed as the standard deviation of the indicator y.

[0070] Step A2: Remove the maximum and minimum values ​​in the biological replicates of the control group and calculate the average value, calculate the ratio of the experimental group data to the control group data, use principal component analysis to reduce the dimension of the fluorescence data, and convert the fluorescence index into a set of linearly unrelated comprehensive indicators;

[0071] In step A2, the maximum and minimum values ​​in the biological replicates of the control group are removed and the average value is calculated, and the data of the experimental group and the data of the control group are compared to eliminate the differences between varieties. The calculation method of the principal component analysis is specifically as follows:

[0072] There are m principal component parameters obtained by principal component analysis, and the degree of explanation of the nth principal component for all indicators is k.

[0073] The calculation method for the ratio of the explanation degree of each principal component to the explanation degree of all principal components for all indicators is:

[0074] .

[0075] Step A3: Establishing a membership function of the comprehensive salt tolerance coefficient by combining the extracted principal components and index weights to obtain the comprehensive salt tolerance coefficient;

[0076] Preferably, , where S represents the membership function, It is expressed as the score of the cth principal component of the dth accession, is expressed as the minimum value of the cth component, Expressed as the maximum value of the cth component;

[0077] The calculation method of comprehensive salt tolerance coefficient is:

[0078] (c) = μ(c)×S(c) (the value of the dth accession after standardization), μ(c) is expressed as the indicator weight;

[0079] , where D represents the comprehensive salt tolerance coefficient and c represents the cth principal component.

[0080] If the calculated comprehensive salt tolerance coefficient is greater than the preset comprehensive salt tolerance coefficient threshold, then the cabbage sample is salt-tolerant; if the calculated comprehensive salt tolerance coefficient is less than the preset comprehensive salt tolerance coefficient threshold, then the cabbage sample is sensitive; if the calculated comprehensive salt tolerance coefficient is equal to the preset comprehensive salt tolerance coefficient threshold, then the cabbage sample is normal; the calculated data are clustered and divided into salt tolerance, sensitivity and normality.

[0081] Step A4: Use MATLAB 2023b software and 4 devices to interface and use Library functions to obtain multispectral image data;

[0082] In step A4, the steps of obtaining multispectral image data are specifically as follows:

[0083] Step A41: 4. The equipment is installed and set up correctly, and MATLAB2023b has been installed on the computer; 4 API or SDK is compatible with MATLAB and tools are available;

[0084] Step A42: Install MATLAB's Image Processing Toolbox;

[0085] Step A43: Use 4. Built-in msi interface; Place all files in a directory called "Multispectral Imaging Toolbox"; Add this directory to the Matlab search path by selecting "Set Path..." in the Matlab menu; Type help msi in the MATLAB command box to get an overview of the functions in the toolbox;

[0086] Step A44: Use the msi function to initialize and 4; set the exposure time and spectral range for multispectral image acquisition; call the msi function to obtain multispectral image data;

[0087] Step A45: Pass The library's functions process, analyze and visualize 4. Acquired data.

[0088] Step A46: Use MATLAB's image processing toolbox to perform image denoising and correction; perform image analysis through the msi library.

[0089] Step A47: Visualize the results through the plotting function of MATLAB to display the processed multispectral image; export the processed data as a file.

[0090] Step A5: Select the near-infrared channel image and perform grayscale processing, and further apply the threshold segmentation method to extract the region of interest; for the extracted ROI area, use the local binary pattern algorithm to convert it into a binary image;

[0091] The method for extracting the region of interest is specifically as follows:

[0092] Step A51: converting the color image into a grayscale image by calculating the average value of the R, G, and B channels of the RGB image;

[0093] Step A52: selecting a threshold to distinguish the target and background in the image;

[0094] The thresholds selected include manually selected fixed values ​​and values ​​automatically determined based on the histogram or other features of the image;

[0095] Common selection methods include Methods,Histogram-based methods;

[0096] Step A53: applying the selected threshold to the grayscale image, dividing the image into two parts: pixels greater than the threshold are targets, and pixels less than the threshold are backgrounds, thereby obtaining a binary image in which the target pixel value is 1 and the background pixel value is 0, and marking the target image as an image of interest;

[0097] Step A54: Perform some post-processing operations on the binary image, use morphological operations to remove noise or connect separated targets; display the processed binary image to check and analyze the segmentation results.

[0098] Step A6: For the same ROI area, calculate its pixel values ​​under 19 multispectral channels and perform inter-channel multiplication operations to generate a new multispectral image of the area;

[0099] In step A6, the calculation method for generating a new multispectral image is specifically as follows:

[0100] , where Z represents the multispectral image value of ROI, Represents the reflectivity of each pixel of the original image. represents the value of each pixel of the binary image, N(k) represents the spectral image of the kth band, and the 19 images are superimposed to obtain a 19-channel ROI multispectral image.

[0101] Step A7: Calculate the average spectral reflectance of the non-zero area of ​​the pixel points in the multispectral image and the corresponding texture characteristics; use the gray level co-occurrence matrix method to analyze the gray level connection characteristics between the pixels on the leaf surface;

[0102] In step A7, the calculation method of the average spectral reflectance is specifically as follows:

[0103] ,in, Expressed as the average spectral reflectance, It is represented as the reflectance of the pixel in the i-th row and j-th column in the n-th band, q is represented as the number of non-zero pixels inside the selected ROI, and m and n represent a total of m rows and n columns respectively.

[0104] The calculation method of color texture features is as follows:

[0105] ,in, Expressed as the first-order moment, N represents the number of pixels, Represented as the i-th color component of the j-th pixel;

[0106] ,in, Expressed as the second-order moment, N represents the number of pixels, Represented as the i-th color component of the j-th pixel, Expressed as first-order moment;

[0107] ,in, Expressed as the third-order moment, N represents the number of pixels, Represented as the i-th color component of the j-th pixel, Expressed as first-order moment;

[0108] Color information is mainly distributed in low-order moments. The first-order moment, second-order moment, and third-order moment express the color distribution of the image. The first-order moment is the mean, which indicates the average intensity of the color component. The second-order moment is the color variance, which indicates unevenness. The third-order moment is the skewness of the color component, which indicates asymmetry.

[0109] The gray level connection characteristic is calculated as follows:

[0110] , where C represents energy and H represents the number of gray levels. It is represented as the image after H level conversion, i is the gray value of the pixel, and j is the gray value of the pixel with a fixed step distance from i;

[0111] , where B represents entropy and H represents the number of gray levels. It is represented as the image after H level conversion, i is the gray value of the pixel, and j is the gray value of the pixel with a fixed step distance from i;

[0112] , where G represents the moment of inertia and H represents the number of gray levels. It is represented as the image after H level conversion, i is the gray value of the pixel, and j is the gray value of the pixel with a fixed step distance from i;

[0113] , where L represents the correlation and H represents the gray level. It is represented as the image after conversion to level H, i is the gray value of the pixel, j is the gray value of the pixel with a fixed step distance from i, It is represented as the average gray value of the corresponding row. It is expressed as the average gray value of the corresponding column, is represented as the variance of the corresponding row, Expressed as the variance of the corresponding column;

[0114] Energy is the sum of the squares of the values ​​of each element in the gray-level co-occurrence matrix, which is used to describe the uniformity or complexity of the image texture. In the gray-level co-occurrence matrix, each element represents the number or probability of co-occurrence between two pixel gray levels, and the energy is the sum of the squares of these co-occurrence numbers or probabilities. The higher the energy value, the more uniform the co-occurrence between the gray levels of pixels in the image, and the more uniform or complex the image texture.

[0115] Entropy is used to describe the characteristics of image texture and quantify the uncertainty or randomness of the grayscale distribution of pixels in the image. In the grayscale co-occurrence matrix, entropy reflects the irregularity of grayscale co-occurrence. When the randomness of the distribution of elements in the grayscale co-occurrence matrix is ​​greater, the entropy value is greater.

[0116] The moment of inertia pulls apart the spatial distribution differences of the image grayscale, reflecting the clarity of the image and the depth of the texture grooves. The deeper the texture grooves, the greater the moment of inertia and the clearer the effect. On the contrary, if the moment of inertia value is small, the grooves are shallow and the effect is blurred.

[0117] The correlation reflects the similarity of the elements of the gray-level co-occurrence matrix in the row or column direction. When the element values ​​are uniformly equal, the larger the correlation value is, and when the element values ​​differ greatly, the smaller the correlation is.

[0118] Step A8: using the gray-level co-occurrence matrix and the local binary method to extract the near-infrared channel grayscale image with obvious characteristics to calculate the texture characteristics of each strain, using the D value as the dependent variable, establishing a multivariate stepwise regression model with the multispectral band parameter as the independent variable, and constructing the optimal regression equation; wherein the multispectral band parameter is the reflectance data of the wavelength of light;

[0119] In step A8, the calculation method of the multispectral texture in the regression equation is specifically as follows:

[0120] ;

[0121] The calculation method of multispectral data in the regression equation is as follows:

[0122] ;

[0123] in, It is related to the chlorophyll content of cabbage. and Related to chlorophyll content, leaf structure and leaf thickness, Related to the water content of leaves, It is related to the pigment content in plant leaves. LBP indicates the concavity and convexity of the cabbage leaf surface and the roughness of the texture. CONTRAST indicates the contrast of the image.

[0124] Step A9: Use the multispectral data collected by VideometerLab4 to construct a Hybrid network model of the framework;

[0125] In step A9, a The hybrid network model of the framework is specifically: including a 1D convolutional neural network and a 2D convolutional neural network, and fusing the features of the two together through a splicing layer; specifically, the 1D-CNN architecture is designed to include a convolutional layer, a normalization layer, and a The activation function layer, a downsampling layer, and an output layer; the 2D convolutional neural network structure includes two convolutional layers, two normalization layers, and a downsampling layer, and also cooperates with activation function, and also has an output layer; then, three fully connected layers are added on top of the fused features, and an activation function is set after them as layer and an output layer for classification tasks to distinguish the target categories; finally, in order to evaluate the performance of the 2D convolutional neural network and splicing network model built based on multispectral images, a five-fold cross-validation strategy was adopted. The accuracy rate and key evaluation indicators F1 value, Precision and Recall obtained by five cross-validations were averaged to obtain the overall performance of the model as the final result measurement standard.

[0126] Among them, 1D-CNN represents 1D convolutional neural network, 2D-CNN represents 2D convolutional neural network, and ROI represents image region of interest.

[0127] It should be noted that the preset values ​​in the present invention are determined according to specific circumstances, that is, this embodiment does not specifically limit the specific values.

[0128] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A salt-tolerant variety identification method based on a combination of multiple indicators and multi-spectral texture, characterized in that: include: Step A1: Use the multifunctional plant photosynthetic phenotype measurement system PAM Time protocol measurement program to measure the complete fluorescence quenching kinetic curve, and use the Pearson correlation coefficient to analyze the correlation between fluorescence data; Step A2: Use principal component analysis to reduce the dimension of the fluorescence data and transform the fluorescence index into a set of linearly unrelated comprehensive indexes; Step A3: Establishing a membership function of the comprehensive salt tolerance coefficient by combining the extracted principal components and index weights to obtain the comprehensive salt tolerance coefficient; , where S represents the membership function, It is expressed as the score of the cth principal component of the dth accession, is expressed as the minimum value of the cth component, Expressed as the maximum value of the cth component; The calculation method of comprehensive salt tolerance coefficient is: (c) = μ(c)×S(c) (the value of the dth accession after standardization), μ(c) is expressed as the indicator weight; , where D represents the comprehensive salt tolerance coefficient and c represents the cth principal component; Step A4: Use MATLAB 2023b software and 4 devices to interface and use Library functions to obtain multispectral image data; Step A5: Select the 850nm near-infrared channel image and perform grayscale processing, and further apply the threshold segmentation method to extract the region of interest; use the OTSU algorithm to convert the extracted ROI area into a binary image; Step A6: For the same ROI area, the binary image is multiplied with the grayscale image of each channel to generate a new multispectral image of the area; Step A7: Calculate the average spectral reflectance of the non-zero area of ​​the pixel points in the multispectral image and the corresponding texture characteristics; use the gray level co-occurrence matrix method to analyze the gray level connection characteristics between the pixels on the leaf surface; Step A8: using the gray-level co-occurrence matrix and the local binary method to extract the near-infrared channel grayscale image with obvious characteristics to calculate the texture characteristics of each strain, using the D value as the dependent variable, establishing a multivariate stepwise regression model with the multispectral band parameter as the independent variable, and constructing the optimal regression equation; wherein the multispectral band parameter is the reflectance data of the wavelength of light; Step A9: Use the multispectral data collected by VideometerLab4 to construct a Hybrid network model of the framework.

2. The salt-tolerant variety identification method based on the combination of multiple indicators and multi-spectral texture according to claim 1 is characterized in that: In step A1, the steps of obtaining a complete fluorescence quenching kinetic curve are specifically as follows: Step A11: preparing the sample Chinese cabbage to be tested; Step A12: measuring the fluorescence index of the sample using the multifunctional photosynthetic phenotyping system; Step A13: Import the raw data measured by the instrument into the Data Analysis software to extract data and images; The calculation method of Pearson correlation coefficient is: , where P represents the Pearson correlation coefficient, x and y represent different fluorescence characteristics, Expressed as the standard deviation of the index x, Expressed as the standard deviation of the indicator y.

3. The salt-tolerant variety identification method based on the combination of multiple indicators and multi-spectral texture according to claim 1 is characterized in that: In step A2, the calculation method of the principal component analysis is specifically as follows: There are m principal component parameters obtained by principal component analysis, and the degree of explanation of the nth principal component for all indicators is k. The calculation method for the ratio of the explanation degree of each principal component to the explanation degree of all principal components for all indicators is: 。 4. The method for identifying salt-tolerant varieties based on the combination of multiple indicators and multi-spectral texture according to claim 1, characterized in that: In step A4, the steps of obtaining multispectral image data are specifically as follows: Step A41:

4. The equipment is installed and set up correctly, and MATLAB 2023b is installed on the computer; 4 API or SDK is compatible with MATLAB and tools are available; Step A42: Install MATLAB's Image Processing Toolbox; Step A43: Use 4 built-in msi interface; Place all files in a directory called "Multispectral Imaging Toolbox"; add this directory to the Matlab search path by selecting "Set Path..." in the Matlab menu; type help msi in the MATLAB command prompt to get an overview of the functions in the toolbox; Step A44: Use the msi function to initialize and 4; set the exposure time and spectral range for multispectral image acquisition; call the msi function to obtain multispectral image data; Step A45: Pass The library's functions process, analyze and visualize 4. Acquired data; Step A46: Use MATLAB's image processing toolbox to perform image denoising and correction; Library for image analysis; Step A47: Visualize the results through the plotting function of MATLAB to display the processed multispectral image; export the processed data as a file.

5. The method for identifying salt-tolerant varieties based on the combination of multiple indicators and multi-spectral texture according to claim 1, characterized in that: The method for extracting the region of interest is specifically as follows: Step A51: converting the color image into a grayscale image by calculating the average value of the R, G, and B channels of the RGB image; Step A52: selecting a threshold to distinguish the target and background in the image; Step A53: applying the selected threshold to the grayscale image, dividing the image into two parts: pixels greater than the threshold are targets, and pixels less than the threshold are backgrounds, thereby obtaining a binary image in which the target pixel value is 1 and the background pixel value is 0, and marking the target image as an image of interest; Step A54: Perform some post-processing operations on the binary image, use morphological operations to remove noise or connect separated targets; display the processed binary image to check and analyze the segmentation results.

6. The method for identifying salt-tolerant varieties based on the combination of multiple indicators and multi-spectral texture according to claim 1, characterized in that: In step A6, the calculation method for generating a new multispectral image is specifically as follows: , where Z represents the multispectral image value of ROI, Represents the reflectivity of each pixel of the original image. represents the value of each pixel of the binary image, N(k) represents the spectral image of the kth band, and the 19 images are superimposed to obtain a 19-channel ROI multispectral image.

7. A salt-tolerant variety identification method based on the combination of multiple indicators and multi-spectral texture according to claim 1: In step A7, the calculation method of the average spectral reflectance is specifically as follows: ,in, Expressed as the average spectral reflectance, It is represented as the reflectance of the pixel in the i-th row and j-th column in the n-th band, q is represented as the number of non-zero pixels inside the selected ROI, and m and n represent a total of m rows and n columns respectively.

8. The method for identifying salt-tolerant varieties based on the combination of multiple indicators and multi-spectral texture according to claim 1, characterized in that: In step A8, the calculation method of multi-spectrum combined with texture in the regression equation is specifically as follows: ； The calculation method of multispectral data in the regression equation is as follows: ； in, It is related to the chlorophyll content of cabbage. and Related to chlorophyll content, leaf structure and leaf thickness, Related to the water content of leaves, It is related to the pigment content in plant leaves. LBP indicates the concavity and convexity of the cabbage leaf surface and the roughness of the texture. CONTRAST indicates the contrast of the image.

9. The method for identifying salt-tolerant varieties based on the combination of multiple indicators and multi-spectral texture according to claim 1, characterized in that: In step A9, a The hybrid network model of the framework is specifically: including a 1D convolutional neural network and a 2D convolutional neural network, and fusing the features of the two together through a splicing layer; specifically, the 1D-CNN architecture is designed to include a convolutional layer, a normalization layer, and a The activation function layer, a downsampling layer, and an output layer; the 2D convolutional neural network structure includes two convolutional layers, two normalization layers, and a downsampling layer, and also cooperates with activation function and also has an output layer.

Images